The paper examines how implementation details—such as data preprocessing, augmentation, and optimization—affect deepfake detector performance more than core architectural choices. By systematically isolating these factors across training, inference, and incremental updates, the authors identify design choices that consistently improve accuracy and generalization. These findings provide architecture‑agnostic best practices that enable state‑of‑the‑art performance on the AI‑GenBench benchmark.
By Lorenzo Pellegrini, Serafino Pandolfini, Davide Maltoni, Matteo Ferrara, Marco Prati, Marco Ramilli
FocusGraph is a framework for selecting keyframes in egocentric long‑video question answering. It uses a lightweight Scene‑Graph LLM Selector to identify query‑relevant clips from graph‑based captions, then extracts keyframes with Patch‑wise Sparse‑Flow Retention (PSFR) before feeding them to a multimodal large language model for answer generation. The method achieves state‑of‑the‑art performance on FindingDory and HourVideo while reducing question‑time inference cost.
By Tatiana Zemskova, Solomon Andryushenko, Ilya Obrubov, Viktoriia Khoruzhaia, Ekaterina Eroshenko, Ekaterina Derevyanka, Dmitry Yudin
Score‑based diffusion models excel at inverse problems but suffer from slow inference and task‑specific hyperparameter tuning. This paper introduces principled reparameterizations that create invariances, enabling the reuse of the same hyperparameters across different tasks. Building on the RED‑diff framework, the authors present OptDiff, a streamlined pipeline that integrates convex optimization to accelerate inference, achieving notable speedups and higher image quality in reconstruction, deblurring, and super‑resolution experiments.
By Julio Oscanoa, Irmak Sivgin, Cagan Alkan, Daniel Ennis, John Pauly, Mert Pilanci, Shreyas Vasanawala
SR‑Ground is a large‑scale dataset created to enable fine‑grained segmentation of visual artifacts in super‑resolved images. It contains 63,000 images processed by various state‑of‑the‑art SR models, each annotated at the pixel level for six distinct artifact types, validated through a crowdsourcing study with 1,062 participants. The dataset improves the training of image quality assessment models with grounding capabilities and supports a fine‑tuning pipeline that reduces perceptible artifacts in SR outputs, outperforming no‑reference methods on both benchmark and real‑world low‑resolution datasets.
By Artem Borisov, Evgeney Bogatyrev, Khaled Abud, Dmitriy Vatolin
VideoWeaver is an agent harness and benchmark for long‑video generation that evaluates and evolves reusable skills. It allows an agent to dynamically compose foundation skills into a workflow based on a single high‑level instruction, rather than following a static pipeline. The benchmark includes 16 task categories and 285 cases with multimodal references, and an evidence‑grounded agent‑as‑judge inspects execution traces and final videos to guide skill evolution, improving both process and output quality.
By Jianhui Wei, Yan Zhang, Jie Tan, Hengchuan Zhu, Xiaotian Zhang, Ziyi Chen, Daoan Zhang, Wei Xu, Yeying Jin, Zuozhu Liu
SlotVLA proposes a slot-attention framework that models both objects and their relations for robotic manipulation, using a relation‑centric decoder and an LLM‑driven action translator. The authors introduce LIBERO+, a fine‑grained benchmark with object‑centric annotations, box and mask labels, and instance‑level temporal tracking to evaluate object‑relation reasoning. Experiments on LIBERO+ show that object‑centric and object‑relation slot representations reduce visual token counts while maintaining competitive generalization, offering a compact and interpretable foundation for multitask visuomotor control.
By Taisei Hanyu, Nhat Chung, Huy Le, Toan Nguyen, Yuki Ikebe, Anthony Gunderman, Duy Nguyen Ho Minh, Khoa Vo, Tung Kieu, Kashu Yamazaki, Chase Rainwater, Anh Nguyen, Ngan Le
DynamicVLA is a latency‑aware Vision‑Language‑Action model designed for dynamic object manipulation, featuring a compact 0.4B architecture and a convolutional vision encoder for efficient multimodal inference. It employs a continuous inference schedule that overlaps reasoning and execution, and a Latent‑aware Action Streaming mechanism that discards stale action prefixes to maintain action‑time alignment. The authors also introduce the Dynamic Object Manipulation (DOM) benchmark, comprising 200K synthetic episodes and 2K real‑world episodes, and demonstrate that DynamicVLA improves dynamic manipulation success in simulation and on real robots.
By Haozhe Xie, Beichen Wen, Jiarui Zheng, Zhaoxi Chen, Fangzhou Hong, Haiwen Diao, Ziwei Liu
ChainLoRA is a replay‑free continual learning framework for large language models that merges task vectors while preserving geometry. It uses chain‑updated training with a one‑sided orthogonality proxy to keep historical state and regularization overhead constant, and applies post‑stream adaptive SVD merging with Procrustes adaptation to separate shared and task‑specific components. Experiments demonstrate state‑of‑the‑art performance on Large and SuperNI benchmarks and competitive results on Standard CL, approaching the scores of replay‑based methods.
By Hang Yin, Haozhe Wang, Yuhua Luo, Zhangqi Pan, Xiaoxing Wang, Junchi Yan
The paper introduces a new algorithm for a nonstationary bandit setting where actions influence both immediate rewards and the evolution of an unobserved latent linear state. By approximating the infinite‑memory reward process with a finite‑memory block‑level proxy and applying a UCB‑based block algorithm, the authors achieve a regret bound of “~O(√T)”, improving upon the previous “~O(T^{2/3})” guarantee. This represents the first such “~O(√T)” result for latent linear‑dynamics bandits with bilinear rewards and an open‑loop action‑sequence benchmark.
By Taehyun Hwang, Hyunjun Choi, Heesang Ann, Min-hwan Oh
VIDA (Visually-Dependent Ambiguity) is a new dataset comprising 2,500 curated instances designed to test whether multimodal machine translation models can use visual input to resolve ambiguous source spans. The authors also introduce Disambiguation-Centric Metrics, which employ an LLM-as-a-judge classifier to verify correct span-level resolution. Experiments with recent large vision‑language models show that visual disambiguation remains difficult, and that chain‑of‑thought supervised fine‑tuning yields stronger out‑of‑distribution performance, especially for collective‑noun ambiguities.
By Jingheng Pan, Xintong Wang, Longyue Wang, Liang Ding, Weihua Luo, Chris Biemann
IQS-BO introduces an in‑context query selection method for Bayesian optimisation that learns to choose evaluation points via supervised learning on synthetic priors. The approach uses a Prior‑data Fitted Network (PFN) to predict, in a single forward pass, the probability that each candidate maximises the objective, thereby amortising the decision step and eliminating the need for costly surrogate refitting or acquisition maximisation. Experiments show that IQS‑BO matches or surpasses traditional Gaussian‑process‑based BO and other in‑context methods on both synthetic and real‑world benchmarks, and that a mixture prior combining GP samples with challenging function types further improves performance.
By Luca Geminiani, Nadja Klein
The paper introduces generalized engression models, a unified nonparametric framework for estimating the full conditional distribution of multivariate outcomes that may include continuous, binary, categorical, ordinal, or ranking variables. Building on the engression deep generative model, it incorporates data‑type‑specific link functions and a stochastic perturbation to smooth the loss, allowing gradient‑based training even with discontinuous links. The authors prove universal representation results for continuous, discrete, and mixed outcomes, and demonstrate that the method matches or surpasses existing type‑specific and joint species distribution models in simulations and two real‑world applications.
By Xinwei Shen, Zijian Guo, Francis Bach
SpikeMoE introduces a spike-based k‑WTA router that uses lateral inhibition and refractory periods to select the top‑K experts based on discrete spike counts, inspired by hippocampal CA1 competition. The framework combines spiking neural network dynamics with mixture‑of‑experts conditional computation and adds a two‑stage missing‑modality module for robust multimodal processing. Experiments on vision, language, and multimodal tasks show that SpikeMoE matches or surpasses ANN baselines while offering energy‑efficient performance.
By Xiaoli Liu, Yujie Liang, Jialin Li, Malu Zhang
The paper introduces the global coherence problem, where AI agents make locally valid decisions that collectively lead to an invalid outcome due to shared state failures. It presents the Observation‑Aliasing Impossibility Theorem, establishing that a policy can guarantee a valid action only when all indistinguishable worlds share an admissible action, and shows that even with additional reasoning, roles, messages, or samples, the missing distinction cannot be recovered. The authors propose a local‑to‑global runtime semantics framework and conduct nine studies demonstrating that missing global state cannot be substituted by local intelligence.
By Xin Heng
The paper evaluates nine on‑device named‑entity recognition models ranging from classical taggers to large language models, measuring not only accuracy but also latency and output validity. Using a silver‑gold benchmark derived from an LLM judge panel and a human‑validated corpus, the study shows that encoder‑based models achieve comparable accuracy to a 4 B instruct LLM while being much smaller, faster, and producing no malformed output. Confidence calibration of GLiNER is analyzed, revealing over‑confidence but improved reliability after temperature scaling and thresholding.
By Vinay Kumar Chaganti
MOMAT (Mixture of Multiple Atlases) is a hardware‑enhanced safety framework designed to defend quantized large language models (qLLMs) on edge devices against jailbreak attacks. It uses a collection of semantic atlases—each containing harmful or benign sample clusters and policy templates—to perform domain‑localized Retrieval‑Augmented Generation. A lightweight Mixture of Experts detector evaluates top‑k similarity features retrieved by a Compute‑in‑Memory (CiM) accelerated engine, achieving a 4.69 × 10⁶‑fold speedup and a 2.5 × 10⁵‑fold energy reduction compared to DRAM‑based baselines while matching state‑of‑the‑art defense performance.
By Boyang Li, Bingyu Shen, Weihao Hong, Zhiyuan Jiang, Xinlei Guan, Yan Ma, Miles Q. Li, Yi Sheng, Ruiyang Qin
SONIC‑O1 is a new benchmark designed to evaluate multimodal large language models on audio‑video understanding. It contains 60 hours of 231 clips across 13 real‑world conversational domains, with 4,958 human‑verified annotations and demographic metadata. The benchmark tests open‑ended summarization, multiple‑choice question answering, and temporally grounded reasoning, revealing performance gaps between model families and across demographic groups.
By Ahmed Y. Radwan, Christos Emmanouilidis, Hina Tabassum, Deval Pandya, Shaina Raza
GENSCRIPT is an inference‑only pipeline that generates synthetic data without training a generative model. It creates a deterministic statistical profile of the source data, uses a language model to infer field semantics and cross‑column constraints, and then compiles these into an executable sampler that works for single‑table, temporal, and relational data. The method builds generators in minutes, samples large datasets quickly, and achieves fidelity comparable to leading methods while preserving key data relationships such as 1‑to‑1 mappings and primary‑foreign key constraints.
By Zilong Zhao, Abdul Raheem, Jiayu Li, Sohei Arisaka, Darius Lim Hong Yi, Milad Abdollahzadeh, Uzair Javaid, Biplab Sikdar
The paper introduces a formal framework for measuring AI propensities—tendencies of models to exhibit particular behaviours—using a bilogistic formulation that identifies an "ideal band" of success probability. It estimates the limits of this band with task‑agnostic rubrics and applies the method to six families of LLMs, showing how shifts in propensity affect task performance. The study finds that propensity estimates from one benchmark predict behaviour on held‑out tasks and that combining propensity with capability metrics yields stronger predictive power than either alone.
By Daniel Romero-Alvarado, Fernando Mart\'inez-Plumed, Lorenzo Pacchiardi, Hugo Save, Siddhesh Milind Pawar, Behzad Mehrbakhsh, Pablo Antonio Moreno Casares, Ben Slater, Paolo Bova, Peter Romero, Zachary R. Tidler, Jonathan Prunty, Luning Sun, Jose Hernandez-Orallo
NeurDuo-EEG is a causal EEG foundation model that introduces channel‑resolved persistent memory and multi‑timescale memory management, allowing it to model continuous EEG with a fixed‑size state. Trained on 3,955 hours of EEG from 17 public datasets, it outperforms baselines on most short‑window and long‑sequence tasks, notably improving seizure detection AUC‑PR from 0.285 to 0.471. The Small variant achieves these gains with only 4.7 M parameters and supports efficient streaming inference with constant per‑chunk latency even as history grows to one hour.
By Yifan Wang, Haiping Liu, Yang Cui, Wenhao Cai, Shuhang Li, Xiaoyang Huang, Xianyang Liu, Jingyu Sun, Yizheng Sun, Cunhang Fan, Tianming Du, Jiancheng Yang, Zhenhong Li, Yunhao Zhang, Hongpeng Zhou, Jingyuan Sun